AlexsJones/llmfit ? reverse-engineered prompt

I want to build a command-line tool that helps people figure out which local LLMs can run on their own hardware. It's confusing to know if a 7B or 70B model will work on your machine, so this app should solve that.

When you run it, it should scan your system's CPU, RAM, and GPU (including VRAM). Then, it should show an interactive list in the terminal with hundreds of popular models. For each model, it should show a score, estimate the performance in tokens/second, and most importantly, give a clear "fit" rating like 'Perfect', 'Good', or 'Won't Run' based on your hardware.

Users should be able to filter this list by use case (like 'coding' or 'chat') and sort by different columns. It should also integrate with common local runtimes like Ollama or llama.cpp, and ideally, have a key to just download the recommended model version. The interface should be a clean, keyboard-navigable TUI.